1. A pluripotency signature predicts histologic transformation and influences survival in follicular lymphoma patients
by Andrew J. Gentles, Ash A. Alizadeh, Su-In Lee, June H. Myklebust, Catherine M. Shachaf, Babak Shahbaba, Ronald Levy, Daphne Koller, and Sylvia K. Plevritis
Blood
Volume 114(15):
October 8, 2009
©2009 by American Society of Hematology

2. Overview of analysis.
Overview of analysis. A module network was constructed from expression data on FL/DLBCL-t7 using Genomica. Modules were annotated by comparison with curated gene sets, and a core network, putatively underlying transformation, was identified that involved differentiation/stem cell–related transcriptional programs. We selected modules for incorporation into a survival model using stepwise regression. The survival model was validated in an independent set from a different group.11 Survival analysis based on core modules was performed with and without proliferation genes.
Andrew J. Gentles et al. Blood 2009;114:
©2009 by American Society of Hematology

3. Modules associated with HT are enriched for ESC-like gene expression signatures.
Modules associated with HT are enriched for ESC-like gene expression signatures. In a sample (column), each module (row) is represented by the mean expression of its component genes (“metagene”). (A) PAM was used to compare FL-pt with DLBCL-t, and modules were ordered by significance, according to their PAM score. Also shown for comparison are the FL-nt samples. There is no coherent differential expression of these modules between transforming and nontransforming FL. (B) Selected associations between transformation-related modules learned from expression data and predefined gene sets as determined by hypergeometric test with permutation correction for multiple hypothesis testing (supplemental Tables 1,4). Except for module 3, FDR values shown here are for gene sets with cell-cycle and proliferation-related genes removed.
Andrew J. Gentles et al. Blood 2009;114:
©2009 by American Society of Hematology

4. A core network of ESC-related modules defines expression changes in HT
A core network of ESC-related modules defines expression changes in HT. (A) A core network representing HT was constructed by extracting 19 modules that distinguished DLBCL-t from FL-pt from the full network of 200 modules.
A core network of ESC-related modules defines expression changes in HT. (A) A core network representing HT was constructed by extracting 19 modules that distinguished DLBCL-t from FL-pt from the full network of 200 modules. Nodes in red indicate modules that are typically up-regulated in DLBCL-T, whereas blue nodes are modules more highly expressed in FL-pt. Stronger coloring and larger size of modules represent the mean fold change in expression between FL-pt and DLBCL-T. White nodes are modules that were not part of the 19-module classifier but are connected to them in the network. Unnumbered modules did not have a significant gene set annotation. Module 9 distinguishes FL-pt from DLBCL-T but is not directly connected to the other core modules in the network. (B) Fold changes for expression of each of the 19 core HT modules were obtained by comparing 12 paired (before and after HT) patient samples in the Glas et al dataset.7 ESC indicates embryonic stem cell signature; HSC, hematopoietic stem cell; and SSS, stromal stem cell.
Andrew J. Gentles et al. Blood 2009;114:
©2009 by American Society of Hematology

5. ESC1 module and MYC show similar expression variation, except in normal centroblasts/centrocytes.
ESC1 module and MYC show similar expression variation, except in normal centroblasts/centrocytes. Expression of the core ESC1 module and MYC (a component of ESC1) was mapped to independent datasets via gene symbols and compared across normal and malignant B cells. Each row represents a separate expression dataset, and the 2 columns show mean expression of ESC1 genes (left, white) and MYC (right, gray). (A-B) Variation across stages of normal B-cell development from bone marrow and tonsil samples,22 showing high expression of ESC1 module but low expression of MYC in CBs/CCs relative to other normal B cells. (C-D) Normal (right of dotted line) and malignant (left) B-cell expression.37 ESC1 module is less highly expressed in FL than in de novo DLBCL (dDLBCL) and normal CBs/CCs, whereas MYC is higher in dDLBCL than in FL or normal CBs/CCs. HSC indicates hematopoietic stem cell; EB, early B cell; ProB, Pro-B cell; PreB, Pre-B cell; ImB, immature B cell; NB, naive B cell; MB, memory B cell; CLL, chronic lymphocytic leukemia; HCL, hairy cell leukemia; MCL, mantle cell lymphoma; FL, follicular lymphoma; dDL, de novo DLBCL; PEL, primary effusion lymphoma; and BL, Burkitt lymphoma.
Andrew J. Gentles et al. Blood 2009;114:
©2009 by American Society of Hematology

6. Core HT module correlates with MYC-driven tumor growth in transgenic FVB/N mouse.
Core HT module correlates with MYC-driven tumor growth in transgenic FVB/N mouse. Gene expression analysis of MYC and expression of ESC1 module in tumor cells from FVB/N mice with conditional overexpression of human MYC that were treated with different doses of doxycycline to down-regulate MYC. At 0.04 to 0.05 ng/mL doxycycline, the level of MYC has dropped below the threshold required for tumor maintenance. (A) The difference between the proportion of tumor cells undergoing growth versus apoptosis shifts in favor of the latter at a specific MYC level (threshold) required to maintain a tumor phenotype. (B) Mean expression of genes in the core FL transformation module, as determined from microarray analysis of mouse tumors at each level of doxycycline.
Andrew J. Gentles et al. Blood 2009;114:
©2009 by American Society of Hematology

7. Three-module stemness model stratifies survival in FL and DLBCL-t and predicts transformation.
Three-module stemness model stratifies survival in FL and DLBCL-t and predicts transformation. A linear predictor score (LPS) based on 3 stemness modules was defined that predicted survival across FL patients in the training set.7 (A) Training set FL patients were split into high-risk/low-risk groups according to whether their LPS was more or less than its median value. (B) Patients with high LPS were more likely to have subsequent transformation of their FL to DLBCL-t, compared with patients with low LPS. (C-D) LPS stratified survival within FL-pt, but not within FL-nt. (E) LPS performance in validation set of 187 FL patients.11 (F) Surprisingly, the LPS defined in FL also stratified patient survival when applied to DLBCL-t gene expression, despite survival times being from initial diagnosis of FL, not time of DLBCL-t biopsy.
Andrew J. Gentles et al. Blood 2009;114:
©2009 by American Society of Hematology